Process capability analysis and estimation scheme for autocorrelated data

<正>Autocorrelation is prevalent in continuous production processes,such as the processes in the chemical and pharmaceutical industries.With the development of measurement technology and data acquisition technology,sampling frequency is getting higher and the existence of autocorrelation cannot be ignored.This paper analyzes five estimation schemes of process capability for autocorrelated data.Comparisons among these schemes are discussed for small sample and large sample.In conclusion,this paper gives a procedure of process capability analysis for autocorrelated data.     

Clustering patents using non-exhaustive overlaps

Patent documents are unique external sources of information that reveal the core technology underlying new inventions.Patents also serve as a strategic data source that can be mined to discover state-of-the-art technical development and subsequently help guide R&D investments.This research incorporates an ontology schema to extract and represent patent concepts.A clustering algorithm with non-exhaustive overlaps is proposed to overcome deficiencies with exhaustive clustering methods used in patent mining and technology discovery.The non-exhaustive clustering approach allows for the clustering of patent documents with overlapping technical findings and claims,a feature that enables the grouping of patents that define related key innovations.Legal advisors can use this approach to study potential cases of patent infringement or devise strategies to avoid litigation.The case study demonstrates the use of non-exhaustive overlaps algorithm by clustering US and Japan radio frequency identification (RFID) patents and by analyzing the legal implications of automated discovery of patent infringement.     

Connecting informal networks to management of tacit knowledge

The relationship between tacit knowledge and informal networks has not been fully explored.Tacit knowledge cannot be well managed directly,because it is deeply embedded in individual actions and experiences.This study proposes an approach to managing informal networks as a new middle way to facilitate the sharing and transferring of tacit knowledge,owing to the fact that tacit knowledge is mostly shared and transferred through informal networks.To support the idea of the approach,an empirical case study of applying the approach to management of a large scientific project is also included.Our findings suggest that informal networks play essential roles for management of tacit knowledge,and thus good management of informal networks can lead to efficient sharing and transferring of tacit knowledge.     

评价决策单元相对效率的修正DEA方法

在原有理论[1] 的基础上提出了一个新的DEA模型———CCRH ,给出了强有效的概念 ,论证了相关理论。CCRH模型与CCR模型相比有以下几个优点 :①描述决策单元的相对效率更确切 ;②利用参考决策单元的生产前沿面来描述评价决策单元“生产”情况得到的结果更加符合实际 ;③可以考查评价决策单元输入、输出变化时对其效率的影响。     

实物期权视角的CAPM有效性检验

实物期权对资产定价的非线性影响会导致本质为线性定价的CAPM模型失效.本文以沪深A股2000-2014年间1503家上市公司为样本,利用市值规模、账面市值比、资本支出和总资产回报四个企业特征变量作为实物期权的代理变量,分别对个股贝塔和预期超额收益率进行实物期权调整,给出了经实物期权调整后CAPM有效的证据;而且,较之市值规模和账面市值比,资本支出和总资产回报更加能够反映实物期权对CAPM有效性的影响.此外,实物期权调整对CAPM有效性的影响对处于生命周期早期阶段的企业、高科技行业以及对股权分置改革和次贷危机后的子样本更为明显.     

不良贷款有无回收判别:一类可选变量的支持向量机方法

针对不良贷款有无回收判别问题中属性变量数目多、示性变量比例高的特点,提出了一类可选择变量的支持向量机方法进行判别预测.该方法一是将逐步回归的支持向量机思想应用在模型变量的选择上,二是将线性逐步回归的结果作为模型选择变量的初始状态,解决了传统支持向量机只能使用固定变量的问题. 实证结果显示,该方法不仅提高了样本外预测的正确率, 而且具有很好的稳健性.     

基于结构分量中目标区域保护的SAR图像压缩

针对无人机载SAR图像高倍率压缩问题,提出了一种基于结构分量中敏感目标区域自动提取与保护的SAR图像压缩策略。首先将SAR图像分解为代表大尺度平滑特征的结构分量和代表小尺度细节纹理特征的纹理分量,然后在结构分量中进行潜在目标区域的检测,生成敏感目标区域掩码,最后对潜在目标区进行保护性的低损压缩,对背景区域进行高损压缩。实验表明,采用本方法的重建图像比标准的JPEG2000算法的重建图像在相同码率条件下具有更好的视觉效果。     

认知无线电中的一种频谱盲感知算法

认知无线电是一种用于提高无线电通信频谱利用率的新的智能技术,它能够通过与其运行环境交互而改变其发射机参数,特别是可让未授权用户使用授权用户的频谱,此外,它还为网络及终端提供更高的灵活性。高可靠性的频谱感知是实现频谱共享的关键。在分析了Parthapratim De提出的认知无线电中频谱Blinding Sensing算法的基础上提出了利用数据前、后向线性预测与矩阵QR分解结合的新算法。仿真实验表明,在同样的无线传输环境下,新算法比原算法能更可靠地感知频谱空穴。     

基于最小二乘支持向量机的交通安全预测模型

分析了最小二乘支持向量机(LS-SVM)在交通安全预测中的优势,确定输入向量集合和输出向量集合,利用LS-SVM建立交通安全预测模型.将1953～2006年全国交通安全相关数据分为训练集和测试集,利用Matlab 7.0进行仿真测试.通过训练LS-SVM得到模型具体参数值,然后对测试集数据进行预测,计算预测误差,并与神经网络模型、SVM模型预测结果进行对比.仿真结果表明,基于LS-SVM建立的交通安全预测模型比神经网络预测模型、SVM模型具有更高的运算速度和预测精确度.     

基于高斯过程回归的大气进入段航天器飞行能力预测方法

轨迹优化技术是目前大气进入段研究的关键技术之一,如何在大气进入动力学复杂、航天器设计参数各异以及进入过程多约束的条件下,对进入轨迹性能参数进行评估是轨迹设计研究的重要问题。对此,以二维落点走廊为表征的大气进入段最大飞行航程作为性能指标,针对传统轨迹优化方法求解计算量庞大的问题,提出了一种基于高斯过程回归(Gaussian process regression, GPR)的大气进入段航天器飞行能力快速预测方法,挖掘航天器进入初始轨迹参量与轨迹包络特征参量之间的映射关系,求解航天器最大航程时避免了复杂的动力学建模以及大规模的迭代寻优过程。利用所提方法对1 000余组不同进入场景的进入轨迹最大航程进行快速预测,将预测结果用于进入段航天器飞行能力评估,为解决大气进入领域相关工程问题提供参考。